Propose: Decision on what PMT type to chose for the TOF detector and
         what strategy to follow.


Facts:  1) We expect about 550 photo electrons from a MIP hitting the
           center of a 252 cm long paddle.

        2) We expect large rates from electromagnetic background for paddles
           close to the beam line.

        3) At 10^8 photons high luminosity running the highest rate for 
           the half width paddle closest to the beam line is 10 MHz.  

        4) The PMTs under consideration are R9779 Hamamatsu 8-stage and
           10-stage tubes. The gain of the 8-stage tube is 5*10^5 at an
           operating voltage of 1500V. The maximum Voltage is 1750V where
           a gain of about 1*10^6 is expected. For the 10-stage tube the 
           gain is about 2.5*10^6 at 1750V with maximum Voltage at 2000V
           where a gain of about 6*10^6 is expected.

        5) For both PMTs a maximum operating current of 0.1mA is given
           which results in a drop in gain of 50% after 1000 hours of
           operation. (That is 42 days.) There are no long term operation
           studies and measurements available from Hamamatsu at this time.


Calculations:

        1) For the highest rate paddle at the center the electromagnetic
           background will cause signals with a mean of about 70% of a MIP
           so about 0.7*550 photo electrons. With a gain for 10^6 and a
           rate of 1MHz (10^7 photons low intensity running) the PMT 
           current is expected to be about 0.062mA. This is lower than
           the maximum recommended operating current and running a 10-stage
           tube at this gain has reasonable headroom to compensate a loss
           in gain with increased high voltage.
           At such a gain the mean signal amplitude for MIPs a the center
           of a TOF paddle is about 350mV to 400mV. With an expected attenuation
           of about 2.3 for hits at the far end of the paddle the smallest
           mean MIP signals are expected to be 152mV (mean) and about 120mV for
           the smallest signals in such a landau distribution. This is still
           reasonable high compared to an assumed discriminator threshold of 
           30mV.
	   This approach will not work at high luminosity with 10^8 photons/s
           in the coherent peak and these high rate PMTs need to be replaced
           by 8-stage tubes described below.

Comment: The 100 microA limit corresponds to 50% degradation in signal in 1000 hours of continuous operation (Do I remember
this right?). For practical running, assume 17 weeks per year continuous operation (35 weeks on the floor at 50% efficiency). 
This is about 2800 hours. Assuming we are willing to run for 1600=1000/0.62 hours, the effective lifetime is only 0.6=1600/2800 
of a year on the floor.
 


        2) Using an 8-stage tube with a gain at 10^5 would require an 
           additional on-board amplifier of 10 to get the signal amplitude to 
           a reasonable hight to have 10^6 total amplification. But the PMT
           current will be a factor of 10 lower than for the 10-stage tube
           example above. The amplifier has to be located on the PMT base and
           has to handle high rates of 10MHz. In fact two amplifiers will be
           required one for the anode output and one for the dynode output
           as outlined below.

        3) It is planed to use a modified PMT assembly with two outputs, a
           negative signal output from the anode and a positive signal from 
           the last dynode capacitively coupled and high impedance terminated.
           This way we do not need any splitter to multiplex the signal for
           ADC and TDC. However this means in the case of the 8-stage tube
           two amplifiers are required one for the anode signal and one
           for the dynode signal and both have to be on the board of the
           voltage divider.

Comment: Note that the dynode signal is about 1/3 the size of the anode and this should be taken into account. Also, if an
amplifier is used, one can easily add two outputs that are relatively isolated from each other and save the expense/complexity
of two amplifiers.


Strategies: There are basically two strategies one can follow


         a) start out with 10-stage PMTs everywhere and run the low luminosity
            period with these tubes and gain experience with this detector.

Comment: I would suggest that a solution directed specifically to the 10^7 rate work for at least 2 years (maybe more). 
>From the numbers you gave above, I am not sure that this is the case. It appears the lifetime is less than one year, unless
I missed something.

            when moving to high luminosity running replace the 10-stage tubes
            with 8-stage tubes for all the paddles that will have rate
            larger than 1MHz. These 8-stage tubes will have special bases
            that incorporate 10-fold amplifiers for the anode and dynode 
            signals.

Comment: If the 10-stage option allows for several years of running, this is an attractive option, because we can 
proceed with it immediately. 


         b) equip the expected high rate paddles with 8-stage tubes right
            from the start where these tubes have special 10-fold fast
            amplifiers for both the dynode and anode signals.